drillr-mcp-server
Financial research data backend for AI agents, offering standardized financial data, semantic news and event search, paragraph-cited company filings, and alternative data across US, Japan, Hong Kong, and China A-shares.
README
<div align="center">
🌐 English · 中文
Drillr · The Financial Research Data Backend for Agents
The financial MCP for AI agents. Scan markets. Build conviction. Track every signal. Cite every claim.
</div>
One API key. Nine tools for agent research: standardized financial data, company discovery, semantic news and event search, paragraph-cited company filings, and alt-data.
⭐ If drillr helps your agent, star us — that's how we know to keep building this in the open.
Quick Start
- Sign up at drillr.ai
- Get an
external-scope API key (formatdrl_xxxxxxxx_xxx..., 45 chars) at drillr.ai/developer/keys — you'll paste this into the config below - Drop into your host's mcp.json — one endpoint exposes all the tools below.
Option A: Manual mcp.json (any MCP host — recommended)
Claude Code / Claude Agent SDK / Cursor / VS Code
{
"mcpServers": {
"drillr": {
"type": "http",
"url": "https://gateway.drillr.ai/mcp/data",
"headers": { "Authorization": "Bearer <YOUR_DRILLR_API_KEY>" }
}
}
}
⚠️ Replace
<YOUR_DRILLR_API_KEY>(including the angle brackets) with thedrl_*key you got in step 2. Power users on Claude Code / Cursor / VS Code can use${DRILLR_API_KEY}instead andexport DRILLR_API_KEY=drl_...in the shell launching the host.
For Cursor, paste the block into ~/.cursor/mcp.json. For VS Code (GitHub Copilot Chat), run MCP: Add Server from the Command Palette and paste the block. Or use one-click install:
After install, replace ${DRILLR_API_KEY} in the generated config with your real drl_* key.
Hermes Agent
mcp_servers:
drillr:
url: 'https://gateway.drillr.ai/mcp/data'
headers: { Authorization: 'Bearer <YOUR_DRILLR_API_KEY>' }
Other hosts
Any MCP-compatible host (OpenClaw, ChatGPT MCP, etc.) — same Streamable HTTP transport. A Bearer API key is the recommended setup. Clients with MCP OAuth support can instead omit the headers block and complete the browser sign-in flow.
Option B: Smithery one-line
npx -y @smithery/cli install drillr/drillr --client claude
Smithery prompts for your drl_* API key on first install and writes it into your client's mcp.json automatically.
Listing: https://smithery.ai/servers/drillr/drillr
Option C: Claude Code plugin
This repo doubles as its own single-plugin marketplace. From Claude Code:
/plugin marketplace add Little-Grebe-Inc/drillr-mcp-server
/plugin install drillr
Then set DRILLR_API_KEY in your environment (or paste it into the generated config) and you're done.
Hello World
Once configured, ask your agent something like:
"Pull NVDA's last 10-Q gross margin and compare it to AMD's same quarter — flag any divergence in segment mix."
What happens under the hood:
- Your host routes the question to the
drillrMCP server - The agent picks the right tools — typically
sec_report_search(10-Q content) andrun_sql(financial_statements for margins) - You get back a markdown answer with sources cited, typically in 8-15 seconds
- Check your remaining credit balance at drillr.ai/developer/keys; REST clients additionally get an inline
{ "data": ..., "_credits": ... }envelope on every 2xx (see REST API › Response Envelope) — MCP responses follow standard JSON-RPC and do not carry per-call credit info inline
One Toolkit, 9 Tools
drillr exposes a single MCP endpoint with 9 tools — an all-in-one toolkit for most financial research workflows:
| Tool | Purpose |
|---|---|
run_sql |
Standardized financial data over 90+ tables — statements, ratios, earnings, insider, ownership, prices, alt-data |
sec_report_search |
Paragraph-level semantic search across company filings in the US, Japan, Hong Kong, and China A-shares |
sec_report_list |
List a ticker's indexed filings by filing type |
company_search |
Four-market qualitative discovery by business model, supply chain, peers, or theme |
news_search |
Semantic search over news, market events, and attributed claims, grouped into storylines |
ticker_lookup |
Resolve a company name, brand, or ticker substring to ticker history |
list_tables |
Discover available alt-data SQL tables by category |
get_table_schema |
Inspect columns and types for any SQL table |
fiscal_utility |
Fiscal-period helpers (FY/FQ resolution across companies with non-calendar years) |
Full tool reference: docs/tools.md.
What's Covered
- Core equity coverage: US, Japan, Hong Kong, and China A-shares. Ticker formats:
AAPL,6758.T,00700.HK,600519.SH/300750.SZ. - Ontology-based Company Search: Search over the universe of equities with business model descriptions, supply chain positions, growth vector or thematic fit.
- Fundamentals:
financial_statements,company_snapshot, andprice_volume_historycover all four core markets; financial history reaches back to the 1980s - Company filings: SEC EDGAR, Japan EDINET, HKEX, and China A-share reports with paragraph-level semantic search
- Earnings: call transcripts with AI-structured summaries and estimate-vs-actuals history; this specialized dataset covers US + Japan
- Markets: equities, ETFs, indices, forex, crypto, commodities
- Specialized US datasets: analyst ratings, ownership, executives, 8-K events, and extended-hours quotes
- News + events: continuously updating four-market and cross-asset search with story grouping and attributed claims
- AI value chain alt-data: energy & power, data centers, semiconductors, compute pricing, AI models / companies / benchmarks, LLM token pricing, macro & trade, prediction markets, critical minerals
Full data dictionary: docs/tools.md.
REST API
Every MCP tool has a 1:1 REST endpoint. Same drl_* key, same data, same billing. See docs/rest-api.md.
curl -X POST https://gateway.drillr.ai/api/v1/data/run_sql \
-H "Authorization: Bearer $DRILLR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"sql":"SELECT ticker, close FROM price_volume_history WHERE ticker='\''AAPL'\'' AND time_frame='\''daily'\'' ORDER BY period_end DESC LIMIT 5"}'
Out of Scope
We're upfront about edges so your agent doesn't waste research loops:
- Private / unlisted companies (we cover public-listed only)
- On-chain crypto metrics — we have CEX prices (BTCUSD / ETHUSD / SOLUSD etc.), not TVL / holders / wallets
- Options chains, real-time order book, intraday tick data
- Retail brokerage actions (placing orders, managing positions)
- drillr does not produce its own price forecasts — we surface analyst consensus
Community
Building something with drillr, hit a rough edge, or want early-access drops? Come say hi — scan to join, or click the heading link.
<table> <tr> <td align="center" width="50%"><a href="https://discord.gg/YAh96nw5Vh"><b>Discord</b></a></td> <td align="center" width="50%"><b>WeChat</b></td> </tr> <tr> <td align="center"><img src="https://gateway.drillr.ai/qr/discord.svg" width="160" alt="Drillr Discord QR" /></td> <td align="center"><img src="https://gateway.drillr.ai/qr/wechat.svg" width="160" alt="Drillr WeChat group QR" /></td> </tr> <tr> <td align="center">Devs building agentic research products — office hours, debugging help, early access.</td> <td align="center">Chinese-speaking dev community — fastest product feedback.</td> </tr> </table>
License
MIT — see LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。